1 23

Telecommunication SystemsModelling, Analysis, Design andManagement ISSN 1018-4864 Telecommun SystDOI 10.1007/s11235-011-9564-7

Localization algorithms of WirelessSensor Networks: a survey

Guangjie Han, Huihui Xu, TrungQ. Duong, Jinfang Jiang & TakahiroHara

1 23

Your article is protected by copyright and

all rights are held exclusively by Springer

Science+Business Media, LLC. This e-offprint

is for personal use only and shall not be self-

archived in electronic repositories. If you

wish to self-archive your work, please use the

accepted author’s version for posting to your

own website or your institution’s repository.

You may further deposit the accepted author’s

version on a funder’s repository at a funder’s

request, provided it is not made publicly

available until 12 months after publication.

Telecommun SystDOI 10.1007/s11235-011-9564-7

Localization algorithms of Wireless Sensor Networks: a survey

Guangjie Han · Huihui Xu · Trung Q. Duong ·Jinfang Jiang · Takahiro Hara

© Springer Science+Business Media, LLC 2011

Abstract In Wireless Sensor Networks (WSNs), localiza-tion is one of the most important technologies since it plays acritical role in many applications, e.g., target tracking. If theusers cannot obtain the accurate location information, therelated applications cannot be accomplished. The main ideain most localization methods is that some deployed nodes(landmarks) with known coordinates (e.g., GPS-equippednodes) transmit beacons with their coordinates in order tohelp other nodes localize themselves. In general, the mainlocalization algorithms are classified into two categories:range-based and range-free. In this paper, we reclassify thelocalization algorithms with a new perspective based on themobility state of landmarks and unknown nodes, and presenta detailed analysis of the representative localization algo-rithms. Moreover, we compare the existing localization al-gorithms and analyze the future research directions for thelocalization algorithms in WSNs.

G. Han (�) · H. Xu · J. JiangDepartment of Information & Communication Systems, HohaiUniversity, Changzhou, Chinae-mail: hanguangjie@gmail.com

H. Xue-mail: xuhuihuiq@gmail.com

J. Jiange-mail: jiangjinfang1989@gmail.com

T.Q. DuongBlekinge Institute of Technology, Karlskrona, Swedene-mail: quang.trung.duong@bth.se

T. HaraDepartment of Multimedia Engineering, Osaka University, Osaka,Japane-mail: hara@ist.osaka-u.ac.jp

Keywords Wireless Sensor Networks · Localizationalgorithms · Mobility · Location accuracy

1 Introduction

The development of MEMS, chip systems and wirelesscommunications technology has fostered, low-powered andmulti-function sensor nodes, which can integrate informa-tion collection, data processing, wireless communicationsand other functions together within the small storage, to gainrapid progress [45]. WSN is a multi-hop self-organizing net-work, where a large number of sensor nodes are deployed.The aim of WSN is to perceive, collect and process the in-formation of sensor nodes within the coverage of the net-work [1].

As a bridge between the physical world and the dig-ital world, WSNs are widely used to deal with sensi-tive information in many fields. Application scenarios ofWSNs include military, industrial, household, medical, ma-rine and other fields, especially in natural disasters moni-toring, early warning, rescuing and other emergency situ-ations. For example, by a smart dust network, suspendednodes in the air space can detect pressure, temperatureand other information of different positions to monitor thequality of the atmosphere. Sensor nodes buried under thebed at different depths can collect temperature, pressureand other data to observe the activity of the glacier [37].Sensor nodes in the birds’ nests can help users to fur-ther research the living habits of birds [36]. In above-mentioned applications, all collected information is basedon the accurate location of sensor nodes. Therefore, local-ization is one of the basic and core technologies in WSNs.Efficient location technology and its optimization meth-ods urgently need to be further resolved in depth. Mak-

Author's personal copy

G. Han et al.

Fig. 1 Localization algorithmsclassification in WSNs

ing a study of the localization technology is very essen-tial for WSNs’ theoretical research and realistic applications[48, 55].

Up to now, the most existing localization algorithmsof WSNs are classified into two categories: range-based[5, 30] and range-free [47, 62]. Range-based techniques usedistance or angle estimates in their locations estimations,while range-free techniques only use connectivity informa-tion between unknown nodes and landmarks. Range-basedtechniques have used Received Signal Strength (RSS) [10],Time of Arrival (TOA) [13], Time Difference of Arrival(TDoA) [9], or Angle of Arrival (AoA) [41]. Landmark nodecan obtain its own location information in advance by GPSsystems or the artificial deployment information. In this pa-per, a node that has got its own location information is calleda landmark. Otherwise, it is an unknown node.

This paper is organized as follow: In Sect. 2, we brieflyintroduce the classification of localization algorithms in anew way. In Sects. 2, 3, 4, 5 and 6, we analyze and summa-rize the typical localization algorithms. In Sect. 7, we dis-cuss and summarize the future research direction for the lo-calization algorithms.

2 Classification of localization algorithms

There has been a large body of research on localization forWSNs over the last few years. The localization process of anunknown node can be described as the node determines itsposition by limited communication with several landmarksusing some specific localization technologies. However, inmany WSNs applications, the unknown node can also cal-culate its position based on the connectivity information

between unknown nodes and landmarks. Currently, thereare four kinds of localization schemes: (1) range-based andrange-free localization algorithms; (2) landmark-based andlandmark-free localization algorithms; (3) fine-grained andcoarse-grained localization algorithms; (4) incremental andconcurrent localization algorithms [14, 23, 54]. The above-mentioned algorithms are classified based on the character-istics of landmarks. However, those classifications are notdistinct enough for further research of the localization al-gorithms without considering the mobility state of sensornodes. Thus, in this paper, we reclassify the localization al-gorithms based on the mobility state of landmarks and un-known nodes, as shown in Fig. 1.

In Fig. 1, localization algorithms are classified into fourcategories: (1) static landmarks, static nodes, (2) static land-marks, mobile nodes, (3) mobile landmarks, static nodesand (4) mobile landmarks, mobile nodes. The common fea-ture of the four categories is that they all need landmarks tolocate the unknown node. In next section, we analyze andsummarize the typical localization algorithms of each cate-gory.

3 Static networks—static landmarks and nodes

Localization algorithms of static landmarks and nodes areapplied in WSNs, where all the nodes are static. The al-gorithms can be classified into two categories according towhether they need physical measurements to obtain distanceor angle information [23]. Localization algorithms within

Author's personal copy

Localization algorithms of Wireless Sensor Networks: a survey

each category can be further classified as range-based orrange-free [14]. Range-based algorithms need to measurethe distances between unknown nodes and landmarks. Then,the measured distances are used to calculate the coordinatesof the unknown node. Range-free algorithms can indirectlyobtain the distances between unknown nodes and landmarksby the connectivity information or the exchanged multi-hoprouting information. Then, the indirectly obtained distancesare used to calculate the coordinates of unknown node. Gen-erally speaking, range-based algorithms can achieve higherlocalization accuracy; however their performance is limitedby the high hardware cost and the heavy power consump-tion. In contrast, range-free algorithms lower the costs andare much more efficient in the localization process of un-known nodes.

3.1 Range-free localization algorithms

Range-free algorithms do not need to measure the distanceor angle information between unknown nodes and land-marks, which estimate the distance between two nodes bythe connectivity information, the energy consuming infor-mation, or the area information of the superimposed regionof the landmarks. In this paper, range-free localization al-gorithms can be divided into four categories: connectivitylocalization algorithms [30, 52], centroid localization algo-rithms [5, 20, 27, 47], energy attenuation localization al-gorithms [44, 46, 58] and region overlap localization algo-rithms [21, 35, 56].

3.1.1 Connectivity localization algorithms

These localization algorithms combine the connectivity ofthe ideological graph theory with the node’s localization.For example, in an undirected graph G, if vertex Vi has pathto connect vertex Vj , then Vi and Vj are connected. If G

is a directed graph, then the paths of Vi and Vj have thesame direction. If any two points of the graph are connected,then G is called a connected graph [3]. The Connectivity lo-calization algorithms of WSNs are closely related to theirnetwork topology [30]. The detailed information of typicalalgorithms [28, 30, 40, 52] is described in the sequel.

In [40], the core idea of DV-Hop is that the nodes ex-change the traditional distance vector packets, so that eachnode has the minimum hops and the coordinates of all land-marks. Then, each landmark broadcasts its average distanceof each hop with data packets. When the unknown node re-ceives the average distance of each hop, it calculates the dis-tance to each landmark according to the recorded hop in-formation. As shown in Fig. 2, the actual distances betweenlandmark i and other two landmarks j and k are dji anddik , respectively. Hop number hij and hik are 3 and 4, re-spectively, then landmark i can calculate the average dis-tance to each hop, which is ci = (dji + dik)/(hij + hik) =

Fig. 2 DV-hop localization algorithm

Fig. 3 Formation of the BN tree

(dji + dik)/(3 + 4). Therefore, the unknown node p obtainsthe average value of each hop to landmark i, and calculatesthe distances to landmarks i, j, k are ci,3ci,4ci respectively.Then p calculates its coordinates using trilateration or max-imum likelihood estimation (MLE) [22]. Simulation resultsshow that DV-Hop does not need the distance to be mea-sured. Compared with the Centroid location algorithm [4],DV-Hop can meet localization requirement in the sparse net-work. When the ratio of landmarks increases, the localiza-tion error of DV-Hop also decreases.

In [30], the researchers propose the concept of Localiz-able Collaborative Body (LCB). LCB is a kind of distance-related methods using the graph theory to implement local-ization. The unknown node using this kind of algorithms canlocate itself with multi-hop landmarks. Therefore, LCB canovercome the restriction of at least three neighbor landmarksneeded to locate the unknown nodes. LCB first needs allthe landmarks to broadcast their location information. Anyunknown node that receives the related location informa-tion can change the model of the network into a BN-tree.In a BN-tree, only the root node has at least three sub-rootnodes, other parent nodes have at least two child nodes. LCBevolved from the BN-tree can complete the localization us-ing the location information of landmarks and the relative lo-cation relationship between unknown nodes and landmarks.

As shown in Fig. 3, nodes 2, 3, 6, 8, 9 are landmarks,and nodes 1, 4, 5, 7 are unknown nodes. The left graph isthe position relationship of WSNs, and the right graph is theBN tree based on graph theory.

LCB can reduce the computation overload and commu-nication cost in the localization process. However, LCB alsohas a shortcoming that can result in cumulative error, thecorresponding cooperative localization between unknownnodes can cause the localization error of one unknown nodeto have great impact on the localization error of other un-known nodes.

Author's personal copy

G. Han et al.

Fig. 4 Tetrahedron method

3.1.2 Centroid localization algorithms

The core idea of the centroid localization algorithms is touse the connectivity relationships among nodes to calculatethe unknown node’s position information. These algorithmsare Range-free localization algorithms. In these algorithms,the landmark periodically broadcasts its coordinates infor-mation, which contains landmark’s ID and location informa-tion, to the neighboring unknown nodes. When the locationinformation received by the unknown node exceeds a cer-tain threshold in a period of time, then the unknown nodeand landmarks can be interconnected. The estimate locationof the unknown node is the centroid of the polygon whichformed by several landmarks. The detailed information oftypical algorithms [5, 6, 20, 26, 47] is described in the se-quel.

The researchers in [5] propose the tetrahedron localiza-tion algorithm. As shown in Fig. 4, nodes A1,A2,A3,A4

are four landmarks. The algorithm calculates the centroid ofeach tetrahedron, and then the average value of the centroidis the estimate position of the unknown node.

The tetrahedron localization algorithm estimates the co-ordinates of the unknown node by calculating the centroidof tetrahedron. Simulation results show the localization er-ror of tetrahedral algorithm is higher than that of traditionalcentroid algorithm [4]. The localization error of tetrahedralalgorithm is 0.54R, and the localization error of traditionalcentroid algorithm is 0.7R. Compared with traditional local-ization algorithms, the localization accuracy of tetrahedronlocalization algorithm improves by 29%. However, the coor-dinates of unknown node need to be calculated many rounds,thus the calculation and the energy consumption are large.

In [47], the algorithm combines DV-Hop with Assump-tion Based Coordinates (ABC) to locate unknown nodes.First, the algorithm calculates the distance between un-known nodes and landmarks by DV-Hop, and then calculatesthe coordinates of the unknown node using ABC algorithm,as shown in Fig. 5, where n0, n1, n2 are landmarks, n3 isan unknown node. Landmark n0 is located in the coordinateorigin, n1 is in Y axis and n2 is in XOY plane. We can cal-culate the coordinates of node n3 based on the geometricrelationship of the landmarks. The main feature of the algo-rithm is that only simple computation is needed. However,the localization error of the algorithm is high when the nodeis irregular placed in a WSN.

Fig. 5 ABC method

Fig. 6 Three-dimensionalcentroid algorithm

The main difference of literature [47] and literature [5]is mathematical method. The former uses ABC method,while the latter applies tetrahedron method. Simulation re-sults show that the average localization error of literature[47] is 0.46R, and that of literature [5] is 0.54R.

A novel three-dimensional centroid algorithm [20] isproposed for unknown nodes in a 3-D WSN, which as-sumes that geometric relationships and communication con-straints between unknown nodes and landmarks are foundedbased on the assistant three-dimensional coordinates sys-tem. The three-dimensional graph is constructed by calcu-lating several profiles and curving planes. In order to de-crease the communication and computational overload, thethree-dimensional graph is converted into a plane graph,which is composed of several profiles. As shown in Fig. 6,dij represents the communication range of landmark Ai ,and θi1, θi2, βi denote the slopes of the three-dimensionalgraph. The plane graph can be constructed by the slopesof three-dimensional graph. Finally, the centroid of planethree-dimensional graph is the estimated position of the un-known node.

Simulation results show that the localization ratio of thealgorithm is close to 99% when landmark density is greaterthan 6. This means that the algorithm has low requirementon landmark density, which is suitably used in sparse net-work environment.

3.1.3 Energy attenuation localization algorithms

When an unknown node is covered by a landmark, the sig-nal amplitude of the landmark becomes smaller as its trans-

Author's personal copy

Localization algorithms of Wireless Sensor Networks: a survey

Fig. 7 Energy grade overlap graph

mission distance increases. And the beacon energy of thelandmark which the unknown node receives also becomessmaller. In energy attenuation localization algorithms, theunknown node calculates the distance with the landmarkbased on the energy attenuation of the beacon. The typicalalgorithms have sound energy attenuation localization algo-rithm [58] and energy range localization algorithm [44, 46].The details of those algorithms are described in the sequel.

In [58], the distance between the unknown and the land-mark can be calculated based on the source energy attenua-tion of the beacon of the landmark. First, the algorithm de-termines the objective’s function by constructing the max-imum likelihood method and then calculates the estimatedposition of the unknown node using Gauss-Newton func-tion. Simulation results show with Signal-to-Noise Ratio(SNR) and signal power increase, the positioning accuracyof the unknown node can be improved greatly. The algo-rithm reflects that the noise has impact on the localization.The localization error of the unknown node decreases from6 m to 2 m, when SNR increases from 36 dB to 45 dB.

In [46], a novel Beacon Signal Ring (BSR) localizationalgorithm is proposed, in which each landmark continu-ously transmits beacons to the unknown nodes with dif-ferent power. The transmission time intervals of two land-marks obey a normal distribution. The algorithm can reducethe interference of two different landmarks. The informa-tion packet of each landmark contains its ID, coordinatesand transmitting power Pi . Each unknown node monitorsand collects the power information of the landmark. The un-known node determines the signal scope based on its re-ceived information. As shown in Fig. 7, landmarks A, B,C transmit signal with different energy grades. The overlapscope of energy (CP ) is the estimated region of the unknownnode. Then the centroid of CP is the estimated coordinatesof the unknown node. The localization accuracy of the algo-rithm is higher than the literature [58]. However, the algo-rithm needs much more energy to transmit different powerbeacons.

A novel energy attenuation localization algorithm forWSNs is presented in [44], which estimates the coordinatesof the unknown nodes based on the Logit normal distribution

Fig. 8 Sketch map of target region

model of energy attenuation. A new “3σ ” principle is pro-posed in the algorithm, which can be used to establish theregion mapping between the received signal power and thetransmission range. With the region mapping technique, anunknown node gets the energy section based on its receivedsignal strengths (RSS) and determines which the transmis-sion range of the landmark is belonged. Then, the algorithmtransforms the region information with distance constraintsto overlap region. The overlapping region is the minimumarea of the unknown node’s position. Finally, the centroidof the region is regarded as the coordinates of the unknownnode. As shown in Fig. 8, nodes B1 and B2 are two land-marks, while node U is an unknown node. Node U esti-mates the energy overlap region based on its received signalstrengths and calculates the estimated coordinates. The lo-calization accuracy of the algorithm is high. However, thealgorithm needs many round calculations, thus the energyconsumption is large.

3.1.4 Region overlap localization algorithms

The core idea of region overlap localization algorithms is tocalculate the centroid of the overlapping region as the esti-mated coordinates of unknown nodes. These algorithms donot need to calculate the distance between unknown nodes,thus reduce network communication overhead and save thenode’s energy consumption. The detailed information oftypical algorithms [21, 35, 56] is listed as follows.

A new HiRLoc localization algorithm is proposed in [21].The algorithm reduces the overlapping region by changingthe transmission power and directional antennas of unknownnodes. As shown in Fig. 9, two nodes L1 and L2 form aregion O(t). The algorithm can obtain the location infor-mation of the unknown node by calculating the centroid ofthe overlap region, in which the landmark’s information in-cludes the coordinates of landmark, the angle of directionalantennas and communication radius R. The traditional algo-rithms rely on the neighbor node’s information to locate un-known nodes, therefore its communication cost increases asthe sensor region increases. However, the algorithm does notdepend on the deployment of landmarks. Simulation resultsshow that when the number of information packets is 15,the localization error can reach 0.2R. In order to achieve the

Author's personal copy

G. Han et al.

Fig. 9 Region overlapping

Fig. 10 Spherical shells overlap

same localization accuracy, APIT algorithm [50] requires200 information packets.

In [35], spherical shell overlap localization algorithm isproposed for WSNs. It divides the whole space into concen-tric spheres with different radii. In particular, a landmarkis located in the center of ball and the distance betweentwo landmarks is the ball radius. The unknown node deter-mines the thinnest layer spherical shell by judging whetherthe sphere contains itself, as shown in Fig. 10. This thinnestlayer spherical shell is the possible region, and its centerof gravity is the estimated location of the unknown node.The advantage of the algorithm is to only require landmarksbroadcasting location information that other unknown nodesuse to locate themselves. Thus, the cost of network commu-nication could be reduced to save the energy consumptionof unknown nodes.

A novel localization algorithm for WSNs is proposedBased on Voronoi graph [56]. The algorithm first sorts re-ceived signal strength receiver (RSSI) of landmarks basedon descending order, and then calculates the Voronoi regionof each landmark using Unit Disk Graph. Then, the centroidof the Voronoi overlap region is regarded as the coordinatesof the unknown node, as shown in Fig. 11. In the processof calculation, all nodes of Voronoi region have the weightvalue defined as the received signal strength value of thelandmark. The location of the node with maximum weightvalue is the estimated location of the unknown node. Sim-ulation results show the localization error of the algorithmdecreases with the increase of the communication radius andlandmark density. When the communication radius and node

Fig. 11 Voronoi graph

density increase, the information of landmarks increases.Thus, the localization accuracy of the unknown node is moreprecise.

3.2 Range-based localization algorithms

Range-based localization algorithms have highly localiza-tion accuracy. But they usually require more hardware in or-der to measure the distance between sensor nodes. The typ-ical distance measurement techniques include RSSI, TOA,TDOA, AOA, and etc. Range-based localization algorithmsuse the above different measurement techniques to locate theunknown node. In this paper, range-based localization algo-rithms can be divided into four categories: bionics localiza-tion algorithms [31, 62], verification localization algorithms[56, 57], landmark placement localization algorithms [11,49] and landmark upgrade localization algorithms [8, 32].

3.2.1 Bionics localization algorithms

The core idea of bionics localization algorithms is to com-bine the model of biological motion with the localizationprocess of unknown nodes. The localization algorithms es-tablish the position relationship between unknown nodesand a landmark according to the laws of biological mo-tion, thus the algorithm can calculate the coordinates of un-known nodes. The detailed information of typical algorithms[31, 62] is described in the sequel.

In [62], the researchers propose a localization algorithmbased on nectar using the living habits of bees in the physicalworld. The algorithm first acquires the distance informationbetween the landmark and the unknown node according tosignal strength, then calculates the relative angle using thecosine law and obtains the relative position of the unknownnodes according to mobility model of bees, finally obtainsthe coordinates of the unknown node. This algorithm has rel-atively high localization accuracy. Simulation results showwhen the distance error is 15%R and the number of land-marks is 20, the localization error can reach 0.14R. How-ever, the algorithm has high requirement for hardware andenergy consumption in the localization process. Thus, it ismore vulnerable to acquired temperature, humidity, obsta-cles and other environmental factors.

Author's personal copy

Localization algorithms of Wireless Sensor Networks: a survey

Fig. 12 Placement of landmarks

3.2.2 Verification localization algorithms

The core idea of these localization algorithms needs to ver-ify the distance value between unknown nodes and land-marks using RSSI value. As a result, the distance measure-ment is the key factor in the localization process. The ap-plications of RSSI are used very widely and fit the actualenvironment in WSNs. The detailed information of typicalalgorithms [56, 57] is described in the sequel.

Wang proposes a weighted RSSI localization algorithmin [57]. Taking into account the impact of environment fac-tors in WSNs, the localization algorithm verifies the weightvalue of a node by the signal strength information and the ac-tual distance. So the algorithm can increase the adaptabilityin different environments and improve the localization accu-racy. Firstly, the algorithm applies RSSI value to calculatethe distance between the nodes. Then, the algorithm usesthe distance and the RSSI value to verify the weight value ofnode. Finally, the algorithm calculates the coordinates of un-known nodes. Simulation results show the probability of thelocalization error of less than 1.5 m is over 85%, while theprobability of the weighted centroid algorithm [56] achievesonly about 75%. Using the same simulation conditions, theaverage localization accuracy can be improved by 25.48%in comparison with that of the weighted centroid algorithm.

3.2.3 Landmark placement localization algorithms

Landmark placement has great relationship with the local-ization error of unknown nodes. Therefore, the geometryrelationship of landmarks is used to improve the localiza-tion accuracy of unknown nodes. The detailed informationof typical algorithms [11, 49] is described in the sequel.

In [11], Han et al. propose Reference Node Selection al-gorithm based on Triangulation (RNST) that the unknownnode’s localization error is the least when three landmarksform an equilateral triangle, as shown in Fig. 12.

Traditional multilateral location algorithm [53] adoptsa random placement and choices several different land-marks to locate the position of unknown nodes. Generally,most sensor nodes work in the resource-constrained envi-ronment with limited computing, storage and communica-tion capabilities. The energy resource is extremely limited.In the localization process, the computation overload is largeand the convergence speed of error is slow. It is difficultfor unknown nodes to achieve its position in a resource-constrained situation. When three landmarks form an equi-lateral triangle, the localization accuracy of RNST decreasesby 34.9% than that of the traditional method [53]. RNST caneffectively improve the localization accuracy of unknownnodes.

3.2.4 Landmark upgrade localization algorithms

Taking the cost of WSNs into account, the number of land-marks is limited as the communication range of each un-known node is difficult to have more than 3 landmarks.However, if the node is to obtain high precision and highcoverage, WSNs need to contain a large number of land-marks. In the case of no increase in the number of land-marks, the algorithms upgrade several unknown nodes tolandmarks. Therefore, the unknown node can receive moreinformation as a landmark and improve the localizationcoverage. The detailed information of typical algorithms[8, 32] is described in the sequel.

A novel landmark upgrade localization algorithm is pro-posed in [8], which upgrades a node with high localizationprecision to a landmark. Unknown nodes are located usingcycle refinement according to the updated location informa-tion of landmarks and control the round number of circu-lation by estimating the variance of coordinate value. Asshown in Fig. 13(a), in the two hop communication rangesof the unknown node U0, the unknown node U2 uses tri-lateration method to determine its position coordinates ac-cording to the distance of three landmarks (L1,L2,L3) tonode U2. Then, node U2 can be upgraded to a landmarkby a number of one- and two-hop landmarks. As shown inFig. 13(b), node U2 is upgraded to a landmark. Node U0

only obtains the distance between two landmarks L1 and L2.Thus, node U0 cannot determine its coordinates. However,in Fig. 13(b), since node U2 is upgraded to a landmark, nodeU0 can get the distance of three landmarks (L1,L2,U2).Therefore, node U0 can use trilateration method to deter-mine its coordinates.

Simulation results show that this algorithm can improvethe localization accuracy with low density of landmarks in

Author's personal copy

G. Han et al.

WSNs [32]. It is significant that the algorithm can reducethe dependence of landmarks. However, the algorithm needsheavy communication and computation. The reason is thatthe algorithm continually sends the updated location infor-mation of landmarks and iteratively computes the coordi-nates of unknown nodes.

3.3 Summary

In contrast with Range-based localization algorithms,Range-free localization algorithms do not need the dis-tance and angle information except the connectivity infor-mation of unknown nodes. The former requires more com-plex equipment and consume heavy computation and com-munication to obtain a relatively accurate location, such ashoney-based localization algorithms [62] and energy attenu-ation localization algorithms [57], etc. Meanwhile, the latterhas gained more and more attention with the advantages ofenergy consumption, such as connectivity localization algo-rithms [30, 52], energy attenuation localization algorithms[44, 46, 58] and region intersection localization algorithms

Fig. 13 (a) Initial position diagram. (b) U2 update as landmark dia-gram

[21, 35, 56], etc. We have analyzed the static landmarks andstatic unknown nodes localization algorithms in the aspectof localization accuracy, node density, landmark density andenergy consumption, as shown in Table 1.

In Table 1, we can see that the localization accuracy ofconnectivity localization algorithms is average and the re-quirement of node density is greater. But the requirement oflandmark density is smaller. On the contrary, the localiza-tion accuracy and energy consumption of energy attenuationlocalization algorithms are good, while their requirementsof node density and landmark density are average. In thecase of region overlap localization algorithms, the localiza-tion accuracy is better; however the requirement of landmarkdensity is higher. The idea of bionics localization algorithmsis novel; however the localization accuracy is average. Ver-ification localization algorithms have average localizationaccuracy and energy consumption, but the requirement oflandmark density is higher. The localization accuracy andenergy consumption of landmark placement localization al-gorithms are good; in contrast, the requirement of landmarkdensity is higher. Finally, landmark upgrade localization al-gorithms can improve position coverage in the case of sparselandmark density, but the localization accuracy is averageand the energy consumption is higher.

4 Static landmarks and mobile nodes

In the real application environment of WSNs, such as moni-toring the living of human or animal, many unknown nodes

Table 1 Comparison of static landmarks and nodes localization algorithms

Localization algorithms Localization Node Landmark Energy

accuracy density density consumption

Connectivity localization algorithms DV-Hop [40] Better Greater Smaller Greater

LCB [30] Average Greater Average Average

Centroid localization algorithms Centroid [5] Average Smaller Greater Greater

ABC [47] Better Smaller Average Average

Three-dimensional centroid [20] Better Smaller Smaller Greater

Energy attenuation localization algorithms Source energy attenuation [58] Average Average Average Average

BSR [46] Better Average Average Greater

Energy intervals [44] Better Average Average Greater

Region overlap localization algorithms HiRLoc [21] Average Smaller Average Average

APIS [35] Better Smaller Greater Smaller

Voronoi [56] Average Smaller Greater Average

Bionics localization algorithms Honey bee orientation [62] Average Average Smaller Greater

Verification localization algorithms weighted centroid algorithm [57] Average Average Greater Average

Landmark placement localization algorithms RNST [11] Better Smaller Average Smaller

Landmark upgrade localization algorithms Landmark sparse [32] Average Smaller Smaller Greater

Author's personal copy

Localization algorithms of Wireless Sensor Networks: a survey

are mobile. The application of mobile unknown nodes isclosely related to our life. Thus, static landmarks and mobilenodes localization algorithms are introduced in this section,which can be divided into historical information localiza-tion algorithms [29, 34, 38, 42] and cluster localization al-gorithms [12]. The former focuses on the historical informa-tion of unknown nodes and the latter focuses on interactionrelationship between landmarks.

4.1 Historical information localization algorithms

The idea of these algorithms is to predict the coordinatesof mobile unknown nodes based on their recorded historicalinformation. The algorithms can save energy computationand are applied to real applications in WSNs. The detailedinformation of typical algorithms [29, 34] is described in thesequel.

A distributed mobile localization algorithm is proposedin [29]. When the number of neighboring landmarks is lessthan 3, the algorithm proposes the idea that unknown nodesuse the predictability of mobility to locate themselves. Inthe algorithm, each unknown node first maintains a historyqueue which saves the information of three latest locations,and then we can obtain the linear motion equations of theunknown node according to the history records. We assumethat the unknown node is a linear motion in short time inter-val and the acceleration is constant. As shown in Fig. 14, thenode ((XA,YA), (XB,YB)) are the points of intersection be-tween tangent L and the circle, and we can acquire the esti-mated coordinates (Xc,Yc) of the unknown node in the nexttime. As a consequence, the coordinates of the unknownnode should is ((XA + XB + XC)/3, (XA + XB + XC)/3).

Fig. 14 Distributed mobile localization algorithm diagram

Fig. 15 DTN localization algorithm

The simulation results show that [29], the positioning cover-age of the algorithm can achieve 99%. When the ranging er-ror increases to 40%, the localization error of the algorithmis only 33% and that of the traditional localization algorithm[61] can reach to 50%.

A Dynamic Triangular (DTN) algorithm, which calcu-lates the coordinates by predicting next RSSI value of theunknown node, is proposed in [34]. As shown in Fig. 15,DTN first finds the possible location of the unknown nodeusing the position information of the landmarks, then cal-culates the estimated distances between the unknown nodeand two landmarks, and estimates the distance measurementerror between the actual distance and the possible distance.Finally, the algorithm selects the coordinates with the leastmeasurement error value as the estimated coordinates of theunknown node. Simulation results show the localization al-gorithm can eliminate the RSSI fluctuation to some extent,thereby improve the localization accuracy of the unknownnodes. The average localization error of the algorithm is1.2 m. However, due to the fact that positioning result hasgreat relationship with the recorded historical informationof the landmarks, thus the accumulative error of the algo-rithm is larger.

4.2 Cluster-based localization algorithms

Cluster-based localization algorithms are suitable for WSNswith low computation complexity. First, these algorithms di-vide a WSN into several clusters, and each landmark respec-tively locates the unknown node in its cluster. Then, the po-sition information of the unknown node is merged in eachcluster. Finally, all unknown nodes can be estimated by thealgorithm. The details information of typical algorithms [12]is listed as follows.

A new distributed target tracking localization algorithmis proposed in [12], which divides a WSN network into sev-eral clusters and each cluster has a landmark which is re-sponsible for finding target, assigning task, and establishingcommunication between clusters. When an unknown nodemoves into a new cluster, the cluster estimates the coor-dinates of the unknown node. If the unknown node movesinto a next cluster, the next cluster is responsible for locat-ing its position. Finally, the mutual cooperation of all clus-ters locates the unknown node. As shown in Fig. 16, eightlandmarks are deployed in the monitoring region, whichbroadcast the beacon to the unknown node and receive itsfeedback beacon. The algorithm calculates the distances be-tween the unknown node and landmarks, and then deter-mines the coordinates of the unknown node using trilater-ation method.

Figure 16 shows that there exists interference when alllandmarks work at the same time. There is possibility that

Author's personal copy

G. Han et al.

Table 2 Comparison of static landmarks and mobile nodes localization algorithms

Localization algorithms Localization Node Landmark Energy

accuracy density density consumption

Historical information Distributed mobile localization Better No effect Smaller Average

localization algorithms algorithm [29] Better No effect Smaller Average

DTN [34] Better No effect Average Greater

Cluster-based Target tracking localization Average No effect Greater Smaller

localization algorithms algorithm [12] Average No effect Greater Smaller

Fig. 16 Target localization algorithm

some beacons can be received by other landmarks with-out the reflection of signal. The received signal of the un-known node may be the reflection of other signals. In orderto avoid the mutual interference, the algorithm makes eightlandmarks working in turn in a very short time. Simulationresults show cluster-based localization algorithm has lowercomputation complexity than the convex programming al-gorithm [53].

4.3 Summary

In a real WSN application, due to the mobility of objectsor users, the position information of mobile nodes is im-portant. In this section, static landmarks and mobile nodeslocalization algorithms can be divided into two categories:historical information localization algorithms and cluster-based localization algorithms. The core idea of literature[29, 38] is based on recorded historical information of mo-bile nodes. The difference is that the former uses varioustransmitting power, while the latter uses the geometric al-gorithm between nodes. Historical information localizationalgorithms consume a lot of energy due to frequently recordproblem. Cluster-based localization algorithms can save en-ergy in some extend. Table 2 is the summary of static land-marks and mobile nodes localization algorithms in the as-pect of localization accuracy, node density, landmark den-sity and energy consumption.

In Table 2, we can draw a conclusion that node densityhas no effect to static landmarks and mobile nodes local-

ization algorithms. The positioning accuracy of historicalinformation localization algorithms is good, but the energyconsumption is higher. The localization accuracy of cluster-based localization algorithms is average and the energy con-sumption is low, but the requirement of landmark density ishigh.

5 Mobile landmarks and static nodes

At present, some localization algorithms use mobile land-marks to locate static nodes according a specific trajectory.The algorithms are divided into two categories: geometriclocalization algorithms [7, 43, 59, 60] and path planning lo-calization algorithms [16–19, 25].

5.1 Geometric localization algorithms

Geometric localization algorithms change the localizationproblem of unknown nodes into a geometry problem, andcalculate the coordinates of the unknown nodes based onthe geometry relationship between mobile landmarks andstatic nodes. The detailed information of typical algorithms[7, 43, 60] is described in the sequel.

In [60], a mobile Location Assistant (LA) according toa specific trajectory periodically broadcasts its position in-formation to unknown nodes. The unknown node calculatesthe distance to the LA using RSSI technique, and then de-termines its own location based on trilateration method, asshown in Fig. 17, LA is a mobile landmark. Simulationresults show that when the distance error is 10% of thecommunication radius, the average localization precision is11.2%. The shortcoming of the algorithm is that it mainlyrelies on the LA equipment. The robustness and security ofLA are the key factors, which needs the precise hardware re-quirement of the LA. The iteration process of the algorithmincreases the network computation and also need to satisfythe requirement of higher storage capacity.

A sphere-based localization algorithm is proposed in [7],which changes the localization problem into the multiplelinear equations to estimate the coordinates of unknown

Author's personal copy

Localization algorithms of Wireless Sensor Networks: a survey

Fig. 17 LA localization algorithm

Fig. 18 Sphere-based localization algorithm

nodes. The algorithm does not need the measurement deviceand other supporting facilities, unknown nodes can only lo-cate themselves through interaction with a mobile landmark.As shown in Fig. 18, the mobile landmark broadcasts the po-sition information at positions A, B, C, D, an unknown nodeE calculates its coordinates based on four received beacons.Simulation results show that when the communication ra-dius of the node is 10 m, the landmark density is 10%, thelocalization error of the unknown node is 50.17%. The al-gorithm uses the least square method to estimate the nodeposition and the filtering strategies method to reduce the lo-calization error. Thus, the positioning accuracy of the un-known node can be improved about 32.12%.

A novel flying landmark localization algorithm is pro-posed in [43], in which each landmark is equipped with aGPS receiver and broadcasts its location information as itflies through the sensing space. Then each unknown nodein the sensing space estimates its own location based on thebasic geometry principles and the received position informa-tion packets from the flying landmark. The unknown nodereceives more than four position information packets, andthen the four positions form two intersecting circles. Thereare two lines through the center of two intersecting circlesand perpendicular to the intersection circle. The intersectionpoints of the two lines are the estimated location of unknownnode. As shown in Fig. 19, A is a mobile landmark, S is an

Fig. 19 Flying landmark localization algorithm

Fig. 20 S shape trajectory

unknown node. In the communication range of the unknownnode, the landmark A moves along the straight line of graph,and periodically broadcasts information packets. Simulationresults show that when the transmission radius is 15 m, thelocalization error of the algorithm is 1.6 m, however the lo-calization error of the centroid algorithm is 2.4 m.

5.2 Path planning localization algorithms

In path planning localization algorithms, a mobile land-mark moves along a specific trajectory and sends informa-tion packets which contain its position information. The un-known node receives the information packet to locate itself.This approach not only can reduce the cost of WSNs, butalso get higher localization accuracy. However, how to findthe optimal path is the basic problem. The detailed infor-mation of typical algorithms [16–19, 25] is described in thesequel.

The researchers in [17] propose that a mobile landmarkmoves in accordance with S shaped trajectory, as shown inFig. 20. The unknown node in sensing region periodicallyreceives the position information of the mobile landmarkto estimate its coordinates. Simulation results show that thetravelling trajectory of the mobile landmark in the algorithmis shorter and the energy consumption is lower. However, theunknown nodes on the edge of sensing area in a WSN can-not be located, due to the fact that it cannot receive adequateposition information from the mobile landmark.

In [18], Dimitrios et al. proposed three different trajec-tories for a mobile landmark, namely SCAN, DOUBLESCAN, and HILBERT, as shown in Fig. 21. The advantagesand disadvantages of the three trajectories are compared in

Author's personal copy

G. Han et al.

Fig. 21 Three different travelling trajectories

Fig. 22 Spiral trajectory

this paper. The travelling trajectory length of SCAN is theshortest, but many collinear beacons (beacon transmittedby the mobile landmark when it moves on a straight line)do not help localization, since the sensor still cannot deter-mine which side of the line, hence at least one non-collinearbeacon is necessary. DOUBLE SCAN increases a trajectoryalong the y-axis direction. This method solves the collinearproblem, but the positioning accuracy of the marginal nodeis low. Therefore, HILBERT increases many turns to solvethis problem. Simulation results show that when the sensingregion is 420 m × 420 m, the travelling trajectory length andlocalization error of SCAN is 3780 m and 0.86 m, respec-tively. Those of DOUBLE SCAN are 4080 m and 0.85 m,respectively. Those of HILBERT are 3840 m and 0.88 m,respectively.

The researchers in [19] propose a Gaussian-Markov algo-rithm to optimize the travelling trajectory. Simulation resultsshow that the localization error is 0.9 m and the localiza-tion error of traditional centroid algorithm is 4.6 m. But thetravelling trajectory length of Gaussian-Markov algorithmis longer and the repeated trajectory is too much, thereforeGaussian-Markov algorithm has higher energy consumptionthan SCAN, DOUBLE SCAN and HILBERT algorithms.

In [16], a mobile landmark moves along a spiral tra-jectory and periodically broadcasts its location informationpackets. When the unknown node receives more than threelocation information packets from the mobile landmark, theaverage value of all coordinates is the estimated position ofthe unknown node. As shown in Fig. 22, the green pointis the position of the information packet and the spiral lineis the travelling trajectory of the mobile landmark. Simu-lation results show that the algorithm effectively solves the

Fig. 23 Backtracking greedy algorithm

Fig. 24 Breadth-first algorithm

collinear problem in the localization process. The localiza-tion accuracy of the algorithm can be up to 0.2R.

The researchers in [25] propose the backtracking greedyalgorithm and the breadth-first algorithm. These algorithmsdynamically adjust the travelling trajectory based on graphtheory. The two algorithms regard a WSN as a connectedundirected graph, thus the path planning problem can beconverted into the spanning tree and traversal problem. Thewhole WSN is described as a connected undirected graph,where the mobile landmark traverses the spanning tree nodeusing the backtracking greedy algorithm and the breadth-first algorithm and to find the best optimal travelling trajec-tory. Figures 23 and 24 are the trajectories of backtrackinggreedy algorithm and breadth-first algorithm, respectively.

Simulation results show that the above two localizationalgorithm can adapt to a WSN, which the nodes are ran-dom deployed on a large scale and efficiently estimate theunknown nodes with high localization accuracy.

5.3 Summary

On the whole, Dimitrios et al. [18] propose SCAN, DOU-BLE SCAN and HILBERT path planning. Huang [17] and

Author's personal copy

Localization algorithms of Wireless Sensor Networks: a survey

Table 3 Comparison of mobile landmarks and static nodes localization algorithms

Localization algorithms Localization Node Trajectory Travelling Energy

accuracy density length speed consumption

Geometric LA [60] Average No affect Shorter Average Smaller

localization Sphere-based algorithm [7] Better Average Average Greater Average

algorithms Flying landmark algorithm [43] Better Average Average Greater Average

Path Planning S shaped trajectory [17] Average Smaller Average Average Smaller

localization SCAN, etc., [18] Better Smaller Shorter Greater Average

algorithms Gauss-Markov trajectory [19] Average Average Longer No affect Greater

Spiral trajectory [16] Better Smaller Average Average Average

Intelligent trajectory [25] Average Greater Longer No affect Average

Hu [16] propose S-shaped and Spiral trajectory method, re-spectively. What they have in common is no matter how theunknown nodes distribute, the mobile landmark moves ac-cording to a specific trajectory. When the unknown nodeis close to the travelling trajectory of the mobile landmark,the localization accuracy is the highest. When the unknownnode is far away from the trajectory of mobile landmark,the localization accuracy is the lowest or even the positioninformation packets of the mobile anchor node cannot bereceived. Li [24] discusses the optimal travelling trajectoryof a mobile landmark. However, due to many factors affect-ing the problem, few principles of selecting the optimal tra-jectory is given. Hongjun [25] estimates the coordinates ofunknown node based on the graph theory, which considersa WSN as a connected undirected graph, and changes thepath planning problem into the graph traversal spanning treeproblem.

In this paper, we have summarized mobile landmarks andstatic nodes localization algorithms in the aspect of local-ization accuracy, node density, landmark density, travellingspeed and energy consumption, as shown in Table 3. FromTable 3, we can see that the localization accuracy of geo-metric localization algorithms is better and the energy con-sumption of the algorithm is also smaller. But there is cer-tain requirement of node density and the travelling speedof the landmark is greater. The localization accuracy of theoptimal trajectory localization algorithm is better and therequirement of node density is lower. However the travel-ling trajectory length is longer. The travelling speed of thelandmark in Gauss-Markov algorithm and intelligent trajec-tory algorithm has no effect on the localization process ofunknown nodes, but the localization accuracy of unknownnodes is average.

6 Mobile landmarks and mobile nodes

Unknown nodes and landmarks are mobile in a WSN. Thelocalization process of these algorithms is complicated, be-

Fig. 25 Self-organizing localization

cause it is of great significance to some special environment.We have divided these algorithms into two categories: time-based localization algorithms [39, 51] and probability distri-bution localization algorithms [20, 40].

6.1 Time-based localization algorithms

These algorithms mainly rely on the continuous movementof landmarks to locate unknown nodes. The idea of thesealgorithms is to calculate the positions of unknown nodesin a very short time interval. The detailed information oftypical algorithms [39] is described in the sequel.

A self-organizing localization algorithm is proposed in[39]. Due to the continuous movement of node mobility, theposition of the unknown node has not changed much. Thealgorithm locates the unknown node in a very short time in-terval. As shown in Fig. 25, the unknown node receives threecoordinate information packets of the mobile landmark in ashort time, and then calculates its coordinates by trilaterationmethod. Simulation results show that the localization errorincreases with the increase of the distance. When the dis-tance between landmarks and unknown nodes is 30 m, thelocalization error of the unknown node is 2.5 m. However,such algorithm requires the travelling speed of the landmarkto be as slow as possible.

Author's personal copy

G. Han et al.

Table 4 Comparison of mobile landmarks and mobile nodes localization algorithms

Localization algorithms Localization Node Landmark Travelling Energy

accuracy density density speed consumption

Time-based localization Self-organizing localization Average Greater Greater Greater Greater

algorithms algorithm [39]

Probability distribution MCL [2] Better Smaller Smaller Average Greater

localization algorithms

6.2 Probability distribution localization algorithms

The core idea of these localization algorithms is that un-known nodes predict their location using the prior distri-bution probability method, in which can reduce the costof a WSN. The detailed information of typical algorithms[2, 15, 33] is described in the sequel.

A dynamic Monte Carlo Localization (MCL) algorithmfor WSNs is proposed in [2], which the node only can de-tect the position information of current neighboring node atsome point. The algorithm contains two stages: predictionand filtration. At the prediction stage, the unknown nodepredicts its estimated location using distributed switchingequipment based on the reserved information and the mo-bile information of the mobile landmark. At the filtrationstage, the unknown node removes the inconsistent informa-tion from the estimated location. When the landmark den-sity is low and the network communication condition is ex-tremely irregular, MCL still can provide accurate localiza-tion. However, when the unknown node cannot acquire itsestimated location, the prediction and filtration stage needto be implemented continually. In addition, once the filtra-tion of the sampling fails, MCL can cause an infinite loop.Simulation results show that MCL algorithm simulates theposterior probability distribution by the discrete sampling.Therefore, the increase in the number of samples can im-prove the localization accuracy, and also cause overload ofthe memory space and computation of MCL. The maximumtravelling speed of nodes is v = 50 m/s. The number of thesamples changes from 1 to 500 and the number of the land-marks is 1. At the initial stage, the localization error de-creases rapidly. The reason is that the number of the sam-ples is too small. When the number of the samples is up to100, the localization error of the unknown nodes is relativelystable.

6.3 Summary

Mobile landmarks and mobile nodes localization algorithmscan be divided into two categories: time-based localiza-tion algorithms and probability distribution localization al-gorithms. The former calculates the coordinates of unknownnodes during one time-slot in the localization process, which

can get the high localization coverage, however it can lead toaccumulation error of the unknown nodes. The latter needscontinuously sample the localization information of land-marks, thus the energy consumption of the unknown nodesis larger. We have summarized mobile landmarks and mo-bile nodes localization algorithms in the aspect of local-ization accuracy, node density, landmark density, travellingspeed and energy consumption, as shown in Table 4. FromTable 4, we can see that the localization accuracy of time-based localization algorithms is average, the requirement forlandmark density and node density is higher, and the en-ergy consumption is greater. The reason is that time-basedlocalization algorithms have great relationship with arrivingtime difference of the landmarks’ localization information.The localization accuracy of probability distribution local-ization algorithm is better, the requirement for landmark andnode density is lower, and the travelling speed of landmarkshas little effect on the localization accuracy of unknownnodes. However, the energy consumption of unknown nodesis greater.

7 Summaries and outlook

In this paper, the details of existing localization algorithmsare analyzed. Localization algorithms are classified into fourcategories: (1) static landmarks, static nodes, (2) static land-marks, mobile nodes, (3) mobile landmarks, static nodes and(4) mobile landmarks, mobile nodes. We have summarizedmobile landmarks and mobile nodes localization algorithmsin the aspect of localization accuracy, localization coverage,localization time, landmark number and energy consump-tion, as shown in Table 5.

From Table 5, we can see that each algorithm has itsown characteristics and none is absolutely the best. On thewhole, mobile landmarks and static nodes localization al-gorithms, such as LA localization algorithm, Sphere-basedlocalization algorithm and Flying landmark localization al-gorithm, can fully prove that the flexibility of mobile nodescan achieve the impossible task of static nodes. Mobile land-marks first periodically broadcast the position informationpackets to unknown nodes, and unknown nodes can estimatetheir positions based on some localization techniques, such

Author's personal copy

Localization algorithms of Wireless Sensor Networks: a survey

Tabl

e5

Com

pari

son

ofdi

ffer

entc

ateg

orie

slo

caliz

atio

nal

gori

thm

s

Loc

aliz

atio

nal

gori

thm

sL

ocal

izat

ion

Loc

aliz

atio

nL

ocal

izat

ion

Lan

dmar

kE

nerg

y

accu

racy

cove

rage

time

num

ber

cons

umpt

ion

Stat

icla

ndm

arks

,C

onne

ctiv

itylo

caliz

atio

nal

gori

thm

sD

V-H

op[4

0]B

ette

rB

ette

rL

onge

r≥

3G

reat

er

stat

icno

des

LC

B[3

0]A

vera

geA

vera

geL

ong

≥3

Ave

rage

Nov

elC

entr

oid

[5]

Ave

rage

Bet

ter

Lon

ger

≥3

Gre

ater

Cen

troi

dlo

caliz

atio

nal

gori

thm

sA

BC

[47]

Bet

ter

Ave

rage

Lon

g≥

3A

vera

ge

Thr

ee-d

imen

sion

alce

ntro

id[2

0]B

ette

rA

vera

geL

onge

r≥

3G

reat

er

Sour

ceen

ergy

atte

nuat

ion

[58]

Ave

rage

Bet

ter

Lon

g≥

3A

vera

ge

Ene

rgy

atte

nuat

ion

loca

lizat

ion

algo

rith

ms

BSR

[46]

Bet

ter

Ave

rage

Lon

ger

≥3

Gre

ater

Ene

rgy

inte

rval

s[4

4]B

ette

rB

ette

rL

onge

r≥

3G

reat

er

HiR

LO

C[2

1]A

vera

geB

ette

rL

ong

≥3

Ave

rage

Reg

ion

over

lap

loca

lizat

ion

algo

rith

ms

API

S[3

5]B

ette

rA

vera

geA

vera

ge≥

3Sm

alle

r

Vor

onoi

[56]

Ave

rage

Bet

ter

Lon

g≥

3A

vera

ge

Bio

nics

loca

lizat

ion

algo

rith

ms

Hon

eybe

eor

ient

atio

n[6

2]A

vera

geA

vera

geL

onge

r≥

3G

reat

er

Ver

ifica

tion

loca

lizat

ion

algo

rith

ms

Wei

ghte

dal

gori

thm

[57]

Ave

rage

Ave

rage

Lon

g≥

3A

vera

ge

Lan

dmar

kpl

acem

enta

lgor

ithm

sR

NST

[11]

Bet

ter

Bet

ter

Ave

rage

≥3

Smal

ler

Lan

dmar

kup

grad

elo

caliz

atio

nal

gori

thm

sSp

arse

land

mar

k[3

2]A

vera

geB

ette

rA

vera

ge≥

3G

reat

er

Stat

icla

ndm

arks

His

tori

cali

nfor

mat

ion

loca

lizat

ion

Dis

trib

uted

mob

ilelo

caliz

atio

nal

gori

thm

[29]

Bet

ter

Low

erL

onge

r≥

2A

vera

ge

mob

ileno

des

algo

rith

ms

DT

N[3

4]B

ette

rA

vera

geL

onge

r≥

3G

reat

er

Clu

ster

-bas

edlo

caliz

atio

nal

gori

thm

sTa

rget

trac

king

loca

lizat

ion

algo

rith

m[1

2]A

vera

geA

vera

geL

onge

r≥

3G

reat

er

LA

[60]

Ave

rage

Ave

rage

Ave

rage

≤2

Smal

ler

Mob

ilela

ndm

arks

Geo

met

ric

loca

lizat

ion

algo

rith

ms

Sphe

re-b

ased

algo

rith

m[7

]B

ette

rA

vera

geL

ong

≤2

Ave

rage

stat

icno

des

Flyi

ngla

ndm

ark

algo

rith

m[ 4

3]B

ette

rB

ette

rL

ong

≤2

Ave

rage

Ssh

aped

traj

ecto

ry[1

7]A

vera

geA

vera

geA

vera

ge1

Smal

ler

SCA

N,e

tc.,

[18]

Bet

ter

Ave

rage

Lon

g1

Ave

rage

Path

plan

ning

loca

lizat

ion

algo

rith

ms

Gau

ss-M

arko

vtr

ajec

tory

[19]

Ave

rage

Ave

rage

Lon

ger

1G

reat

er

Spir

altr

ajec

tory

[16]

Bet

ter

Bet

ter

Lon

g1

Ave

rage

Inte

llige

nttr

ajec

tory

algo

rith

m[2

5]A

vera

geB

ette

rL

ong

1A

vera

ge

Mob

ilela

ndm

arks

Tim

e-ba

sed

loca

lizat

ion

algo

rith

ms

Self

-org

aniz

ing

loca

lizat

ion

algo

rith

m[3

9]A

vera

geA

vera

geL

onge

r≥

3G

reat

er

mob

ileno

des

Prob

abili

tydi

stri

butio

nlo

caliz

atio

nM

CL

[2]

Bet

ter

Ave

rage

Lon

ger

≥3

Gre

ater

algo

rith

ms

Author's personal copy

G. Han et al.

TOA, TDOA, etc. These above algorithms improve the lo-calization accuracy of unknown nodes, and have the charac-teristics of strong distribution, scalability, security and en-ergy efficiency. Thus, a new research category is graduallyformed.

In recent years, solving the localization problem inWSNS has resulted in many innovative solutions and ideas.However, the research in this field is still at the start-upphase. The study has proposed more and more issues. Thefuture research direction of localization algorithms possiblyis static landmarks and static nodes localization algorithmsto be combined with the idea of landmark placement local-ization algorithms and the weighted-based of verification lo-calization algorithms further improve the localization accu-racy of unknown nodes. Static landmarks and mobile nodeslocalization algorithms can be developed based on the issueof landmark placement. Mobile landmarks and static nodeslocalization algorithms can be developed based on the opti-mal trajectory of mobile landmarks. Mobile landmarks andmobile nodes localization algorithms can be combined withthe idea of node tracking. We believe that in addition to theexisting research issues of localization algorithms, the pos-sible hot research topics are: (1) Evaluate the performancemodel of localization algorithms, and improve the landmarkselection and filtering mechanisms to reduce the localizationtime. (2) Randomly deploy the nodes on the surface of theactual land-based, and study the localization performance ofactual land. (3) Find a localization algorithm which is suit-able for resource-constrained sensor nodes, and reduce thelocalization error caused by random distribution of nodes.(4) Research a self-adjustment localization algorithm in themobile network environment, and simulate the localizationalgorithm performance in the low mobility of sensor nodes.(5) Research the optimal path planning in which mobilelandmarks can traverse the entire network. (6) Research thesecurity issue of localization algorithm. (7) Localization is-sue can also be used other environment, such as underwater,underground, body, mobile, and multimedia, which are theinteresting research issues in the future.

Acknowledgements The work is supported by “the Fundamen-tal Research Funds for the Central Universities, No. 2010B22814,2010B22914, 2010B24414” and “the research fund of Jiangsu KeyLaboratory of Power Transmission & Distribution Equipment Technol-ogy, No. 2010JSSPD04”. Takahiro Hara’s research in this paper wassupported by Grant-in-Aid for Scientific Research (S) (21220002) ofthe Ministry of Education, Culture, Sports, Science and Technology,Japan.

References

1. Akyildiz, I.F., & Su, W. (2002). Wireless sensor networks: a sur-vey. Computer Networks, 38(4), 393–422.

2. Baggio, A., & Langendoen, K. (2008). Monte Carlo localizationfor mobile wireless sensor networks. Ad Hoc Networks, 6(5), 718–733.

3. Buckley, F., & Lewinter, M. (2002). A friendly introduction tograph theory (pp. 1–7). New York: Prentice Hall Pearson.

4. Bulusu, N., Heidemann, J., & Estrin, D. (2000). GPS-less low-cost outdoor localization for very small devices. IEEE WirelessCommunications, 7(5), 28–34.

5. Chen, H. (2008). Novel centroid localization algorithm for three-dimensional wireless sensor networks. In Proc. of the 4th interna-tional conference on IEEE wireless communications (pp. 1–4).

6. Chen, W., Li, W., & Heng, S. (2006). Weighted centroid localiza-tion algorithm based on RSSI for wireless sensor networks. Jour-nal of Wu Han University of Technology, 30(2), 265–268.

7. Dai, G., Zhao, C., & Qiu, Y. (2008). A localization scheme basedon sphere for wireless sensor network in 3D. Acta ElectronicsSinica, 36(7), 1297–1303.

8. de Oliveira, H. A. B. F., Nakamura, E. F., & Loureiro, A. A. F.(2005). Directed position estimation: a recursive localization ap-proach for wireless sensor networks. In Proceeding of the 14thinternational conference on communications and networks (pp.557–562).

9. Girod, L., & Estrin, D. (2001). Robust range estimation usingacoustic and multimodal sensing. In Proc. of the IEEE roboticsand automation society (pp. 1312–1320).

10. Girod, L., Bychovskiy, V., Elson, J., & Estrin, D. (2002). Locatingtiny sensors in time and space: a case study. In B. Wemer (Ed.),Proc. of the 2002 IEEE int’l conf. on computer design: VLSI incomputers and processors, Freiburg (pp. 214–219). Los Alamitos:IEEE Computer Society.

11. Han, G., Choi, D., & Lim, W. (2009). Reference node placementand selection algorithm based on trilateration for indoor sensornetworks. Wireless Communications and Mobile Computing, 9,1017–1027.

12. Hao, Y. (2006). Target localization and track based on the energysource. Master’s thesis, Fudan University.

13. Harter, A., Hopper, A., Steggles, P., & Ward, P. (1999). Theanatomy of a context-awe application. In Proc. of the 5th annualACM/IEEE int’l conf. on mobile computing and networking, Seat-tle (pp. 59–68). New York: ACM Press.

14. He, T., Huang, C. D., & Blum, B. M. (2003). Range-free local-ization schemes in large scale sensor networks. In Proceeding ofthe 9th annual international conference on mobile computing andnetworking (MobiCom) (pp. 81–95).

15. Hu, L., & Evans, D. (2004). Localization for mobile sensor net-works. In Proceeding of the 10th annual international conferenceon mobile computing and networking (MobiCom) (pp. 45–57).

16. Hu, Z., Gu, D., & Song, Z. (2008). Localization in wireless sen-sor networks using a mobile anchor node. In Proceedings of the2008 IEEE/ASME international conference on advanced intelli-gent mechatronics (pp. 602–607).

17. Huang, R., & Zaruha Gergety, V. (2007). Static path planning formobile beacons to localize sensor networks. In Proc of IEEE per-vasive computing and communication (pp. 323–330).

18. Koutsonikolas, D., Das, S. M., & Hu, Y. C. (2007). Path planningof mobile landmarks for localization in wireless sensor networks.Computer Communications, 30(13), 577–2592.

19. Kuang, X., Shao, H., & Feng, R. (2008). A new distributed lo-calization scheme for wireless sensor networks. Acta AutomaticaSinica, 34(2), 344–348.

20. Lai, X., Wang, J., & Zeng, G. (2008). Distributed positioning algo-rithm based on centroid of three-dimension graph for wireless sen-sor networks. System Simulation Technology, 20(15), 4104–4111.

21. Lazos, L., & Poovendran, R. (2006). HiRLoc: high-resolution ro-bust localization for wireless sensor networks. IEEE Journal onSelected Areas in Communications, 24(2), 233–246.

22. Li, C. (2006). Wireless sensor network self-positioning technol-ogy. Southwest Jiao tong University, Master Thesis, pp. 15–21.

23. Li, X., Xu, Y., & Ren, F. (2007). Wireless sensor network technol-ogy (pp. 191–218). Beijing: Beijing Institute of Technology Press.

Author's personal copy

Localization algorithms of Wireless Sensor Networks: a survey

24. Li, S., Xu, C., & Yang, Y. (2008). Getting mobile beacon path forsensor localization. Journal of Software, 19(2), 455–467.

25. Li, H., Bu, Y., & Han, X. (2009). Path planning for mobile an-chor node in localization for wireless sensor networks. Journal ofComputer Research and Development 46(1), 129–136.

26. Li, J., Wang, K., & Li, L. (2009). Weighted centroid localizationalgorithm based on intersection of anchor circle for wireless sen-sor network. Journal of Ji Lin University Engineering and Tech-nology, 39(6), 1649–1653.

27. Li, Z., Wei, Z., & Xu, F. (2009). Enhanced centroid localizationalgorithm and performance analysis of wireless sensor networks.Sensors and Actuators 22(4), 563–566.

28. Lin, J., Liu, H., & Li, G. (2009). Study for improved DV-Hop lo-calization algorithm in WSN. Application Research of Computers,26(4), 1272–1274.

29. Liu, Y. (2008). Distributed mobile localization algorithms of WSN.Master’s thesis, Hunan Technology University, pp. 32–35.

30. Liu, K., Wang, S., & Zhang, F. (2005). Efficient localized local-ization algorithm for wireless sensor networks. In Proc. 5th inter-national conference on computer and information technology (pp.21–23).

31. Liu, T., Chen, H., & Li, G. (2009). Biomimetic robot fish self-positioning based on Angle and acceleration sensor technology. InRobot technology and application (pp. 2–6).

32. Liu, M., Wang, T., & Zhou, Z. (2009). Self-localization algorithmfor sensor networks of sparse anchors. Computer Engineering,35(22), 119–121.

33. Lu, K., Zhang, J., & Wang, G. (2007). Localization for mobilenode based on sequential Monte Carlo. Journal of Beijing Univer-sity of Aeronautics and Astronautics, 33(8), 886–889.

34. Luo, R. C., Chen, O., & Pan, S. H. (2005). Mobile user local-ization in wireless network using grey prediction method. In The32nd annual conference of IEEE industrial electronics society (pp.2680–2685).

35. Lv, L., Cao, Y., Gao, X., & Luo, H. (2006). Three dimensional lo-calization schemes based on sphere intersections in wireless sen-sor network (pp. 48–51). Beijing: Beijing Posts and Telecommu-nications University.

36. Mainwaring, A., Polluters, J., & Szewczyk, R. (2002). Wirelessnetworks for habitat monitoring. In Proceeding of the lst ACM in-ternational workshop on wireless sensor networks and applica-tions (pp. 88–97).

37. Martninez, K., Hart, J. K., & Stdanov, J. (2004). A sensorweb for glaciers. In Proc. European workshop sensor networks(EWSN’04), Berlin, Germany (pp. 1–4).

38. Meng, Z., & Song, B. (2009). HWC localization algorithm ofwireless sensor network. Computer Engineering, 35(7), 104–109.

39. Neuwinger, B., Witkowski, U., & Ruckert, U. (2009). Ad-hoccommunication and localization system for mobile robots. In Lec-ture notes in computer science (pp. 220–229). Berlin: Springer.

40. Niculescu, D., & Nath, B. (2003). DV based positioning in ad hocnetworks. Journal of Telecommunication Systems, 22(14), 267–280.

41. Niculescu, D., & Nath, B. (2003). Ad hoc positioning system(APS) using AoA. In Proc. of the IEEE computer and commu-nications societies (pp. 17–34).

42. Ogawa, T., Yoshino, S., Shimizu, M., & Suda, H. (2003). A newin-door location detection method adopting learning algorithms.In Proc. of the 1st IEEE international conference on pervasivecomputing and communications (pp. 525–530).

43. Ou, C., & Ssu, K. (2008). Sensor position determination with fly-ing anchors in three-dimensional wireless sensor networks. IEEETransactions on Mobile Computing, 7(9), 1084–1097.

44. Qin, W., Feng, Y., & Zhang, X.-T. (2009). Localization algorithmfor wireless sensor network based on characteristics of energy at-tenuation. Journal of Chinese Computer Systems, 30(6), 1082–1088.

45. Qiu, Y., Zhao, C., & Dai, G. (2008). Wireless sensor network nodeposition technology. Computer Science 35(5), 47–50.

46. Rui, L. (2008). Underwater GPS location technology. Xi’an Elec-tronic Science and Technology University, Master’s thesis (pp.47–51).

47. Shu, J., Liu, L., & Chen, Y. (2009). A novel three-dimensionallocalization algorithm in wireless sensor networks, wireless com-munications, networking and mobile computing. In Proc. 5th in-ternational conference on wireless communications (pp. 24–29).

48. Sun, L., Li, J., & Yu, C. (2005). Wireless sensor network (pp. 135–150). Beijing: Tsinghua University Press.

49. Sun, P., Zhao, H., & Han, G. (2007). Chaos triangle compliantlocalization reference node selection algorithm. Journal of Com-puter Research and Development, 44(12), 1987–1995.

50. Tian, H., Chengdu, H., & Blum, B. M. (2003). Range free localiza-tion schemes for large scale sensor networks. In The annual inter-national conference on mobile computing and networking (pp. 81–95).

51. Uchiyama, A., Fuji, S., & Maeda, K. (2007). Ad-hoc localizationin urban district. In IEEE international conference on computercommunications (pp. 2306–2310).

52. Wang, D. (2006). Localization of wireless sensor network. South-west Jiao tong University, Master Thesis (pp. 37–45).

53. Wang, F., Shi, L., & Ren, F. (2005). Self-localization systemsand algorithms for wireless sensor networks. Journal of Software,16(5), 858–859.

54. Wang, S., Hu, F., & Qu, X. (2007). Wireless sensor networks the-ory and applications (pp. 142–164). Beijing: Beijing Universityof Aeronautics and Astronautics Press.

55. Wang, G., Tian, W., & Jia, W. (2007). Location update based on lo-cal routing protocol for wireless sensor networks. High-tech Com-munications, 17(6), 563–568.

56. Wang, J., Huang, L., & Xu, H. (2008). A novel range free lo-calization scheme based on Voronoi diagrams in wireless sensornetworks. Journal of Computer Research and Development 45(1),119–125.

57. Wang, Y., Huang, L., & Xiao, M. (2009). Localization algorithmfor wireless sensor network based on RSSI verification. Journal ofChinese Computer Systems, 30(1), 59–62.

58. Yu, H., Chen, X., & Fan, J. (2007). Gauss-Newton method basedon energy target localization. Computer Engineering and Applica-tions, 43(27), 124–126.

59. Yu, G., Yu, F., & Feng, L. (2008). A three dimensional localizationalgorithm using a mobile anchor node under wireless channel. InInternational joint conference on neural networks (pp. 477–483).

60. Zhang, L., Zhou, X., & Cheng, Q. (2006). Landscape-3D: a robustlocalization scheme for sensor networks over complex 3D terrains.In Proceedings of 31st annual IEEE conference on local computernetworks (LCN) (pp. 239–246).

61. Zhao, H., Feng, Y., Luo, J., & Yang, K. (2007). Mobile node lo-calization algorithm of wireless sensor network. Journal of HunanUniversity of Science and Technology, 34(8), 74–77.

62. Zheng, S., Kai, L., & Zheng, Z. H. (2008). Three dimensionallocalization algorithm based on nectar source localization modelin wireless sensor network. Application Research of Computers,25(8), 2512–2513.

Author's personal copy

G. Han et al.

Guangjie Han is currently an As-sociate Professor of Department ofInformation & Communication Sys-tem at Hohai University, China. Heis also a visiting research scholarof Osaka University from Oct. 2010to Oct. 2011. He finished the workas a Post doctor of Department ofComputer Science at Chonnam Na-tional University, Korea, in Febru-ary 2008. He received his Ph.D. de-gree in Department of ComputerScience from Northeastern Univer-sity, Shenyang, China, in 2004. He

has published over 80 papers in related international conferences andjournals. He has served as an editor of ANC and IJIDCS. He has servedas a Co-chair for more than 10 international conferences/workshops;a TPC member of more than 30 conferences. He has served as a re-viewer of more than 20 journals. His current research interests aresecurity and trust management, localization and tracking, cooperativecomputing for Wireless Sensor Networks. He is a member of IEEE.

Huihui Xu is currently pursuingMaster degree from Departmentof Information & CommunicationEngineering at Hohai University,China. She received her B.S. degreein Information & CommunicationEngineering from Hohai University,China, in 2008. Her current researchinterests are localization and track-ing for Wireless Sensor Networks.

Trung Q. Duong was born in HoiAntown, Quang Nam province, Viet-nam, in 1979. He received the B.S.degree in electrical engineering fromHo Chi Minh City University ofTechnology, Ho Chi Minh City,Vietnam, in 2002 and the M.Sc. de-gree in computer engineering fromKyung Hee University, Suwon, Ko-rea, in 2005. He is currently work-ing toward the Ph.D. degree in ra-dio communications with BlekingeInstitute of Technology, Karlskrona,Sweden. In April 2004, he was with

the Faculty of Electrical Engineering, Ho Chi Minh City Universityof Transport, Ho Chi Minh City, Vietnam. In December 2007, he wasa Research Staff with the Radio Communications Group, BlekingeInstitute of Technology. He was a Visiting Scholar with the Polytech-nic Institute of New York University, Brooklyn, USA, from December

2009 to January 2010. He is the author or co-author of some 50 tech-nical papers in the area of wireless and mobile communications. Hiscurrent research interests include cross layer optimization, cooperativecommunications, cognitive radio networks, and applied game theoryin wireless systems. Mr. Duong has been acting as a Reviewer fornumerous journals/conferences and a Technical Program Committee(TPC) member for more than ten conferences. He is a TPC Co-chairof the 2011 International Workshop on Multimedia Streaming (Mul-tiStreams2011). He was the recipient of the Korean Government ITScholarship Program for International Graduate Students from 2003to 2007. He received the Best Paper Award of the IEEE Student Pa-per Contest-IEEE Seoul Section, in December 2006 and was listedas a finalist for the best paper award at the IEEE Radio and WirelessSymposium, San Diego, CA, in 2009.

Jinfang Jiang is currently pursu-ing Master degree from Departmentof Information & CommunicationEngineering at Hohai University,China. She received her B.S. degreein Information & CommunicationEngineering from Hohai University,China, in 2009. Her current researchinterests are security and localiza-tion for Wireless Sensor Networks.

Takahiro Hara received the B.E,M.E, and Dr.E. degrees from OsakaUniversity, Osaka, Japan, in 1995,1997, and 2000, respectively. Cur-rently, he is an Associate Professorof the Department of MultimediaEngineering, Osaka University. Hehas published more than 100 inter-national Journal and conference pa-pers in the areas of databases, mo-bile computing, peer-to-peer sys-tems, WWW, and wireless network-ing. He has served as a ProgramChair of IEEE International Con-

ference on Mobile Data Management (MDM’06 and 10) and IEEEInternational Conference on Advanced Information Networking andApplications (AINA’09). He was a guest editor IEEE Journal on Se-lected Areas in Communications, Sp. Issues on Peer-to-Peer Commu-nications and Applications. He has served as a TPC member of morethan 120 international conferences such as IEEE ICNP, WWW, DAS-FAA, ACM MobiHoc, and ACM SAC. His research interests includedistributed databases, peer-to-peer systems, mobile networks, and mo-bile computing systems. He is an IEEE Senior member and a memberof four other learned societies including ACM.

Author's personal copy